The problem
Early interview rounds take a lot of interviewer time and are hard to keep consistent across large candidate pools. The goal was to reduce that dependence on live interviewers for defined interview stages.
What the system does
- Runs multi-turn interviews as a conversation, not a fixed questionnaire.
- Generates questions and follow-ups dynamically from the candidate’s answers.
- Supports coding questions through an integrated code editor.
- Works in natural-language and video interview formats.
- Includes identity and impersonation controls, which surface signals for human review.
- Produces LLM-generated interview insights for the people making the hiring decision.
My role
I architected the agentic interview flow and led the system’s productisation with the team.